Abstract
The Tibetan Plateau holds 57.2% of China’s lake area and serves as a vital freshwater source for millions in Asia. Its remote lakes, minimally impacted by direct human activity, are sensitive indicators of climate change, with accelerated glacial melt and shifting precipitation patterns posing significant threats to water resources. This study integrates Sentinel-3A satellite altimetry with the Subwaveform Retracking Method and ERA5 climate models to track historical lake level changes. To predict future hydrological shifts, we developed deep learning neural network models for each lake, utilizing CMIP6 climate projections (RCP 4.5 and 8.5) and regional climate data. These neural networks allowed us to generate accurate, lake-specific forecasts of water level dynamics through the 21st century. Key findings reveal that 9 out of 10 lakes exhibit upward trends in water levels, with an average increase of approximately 0.3569 m/yr. The most significant rises were observed in Migriggyangzham Lake (+0.5259 m/yr) and Lexie Wudan Lake (+0.4895 m/yr), while Langacuo Lake showed a decline (-0.2404 m/year). The findings underscore the role of runoff from glacial melt as a consistent driver of rising water levels in northern lakes, whereas eastern lakes are mainly influenced by precipitation and runoff. In contrast, western lakes show higher sensitivity to temperature-induced glacial melt. This is the first large-scale application of deep learning to predict future lake levels on the Tibetan Plateau. By integrating historical and future climate data, this study provides key insights for water resource management and policy development to mitigate the impacts of climate change.
Presenters
Atefeh GholamiPh.D. Candidate, Institute of Atmospheric Physics, University of Chinese Academy of Sciences, Beijing, China
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
Assessing Impacts in Diverse Ecosystems
KEYWORDS
Water Levels, Remote Sensing, SAR Altimetry, Climate Change, Machine Learning